WO2011052602A1

WO2011052602A1 - Ultrasonic imaging device, ultrasonic imaging method and program for ultrasonic imaging

Info

Publication number: WO2011052602A1
Application number: PCT/JP2010/068988
Authority: WO
Inventors: 裕也増井; 東　隆
Original assignee: 株式会社日立メディコ
Priority date: 2009-10-27
Filing date: 2010-10-26
Publication date: 2011-05-05
Also published as: JPWO2011052602A1; EP2494924A1; US20120224759A1; JP5587332B2; CN102596050A; CN102596050B; US8867813B2

Abstract

Provided is an ultrasonic imaging device capable of identifying a noise region in which the echo signal is weak. A reference frame and a comparison frame are selected from an image having two or more frames obtained by processing a received signal. A region of interest is set in the reference frame, a search region wider than the region of interest is set in the comparison frame, a plurality of candidate regions that are candidates for the destination of movement of the region of interest are set in the search region, the similarity between image characteristic values within the region of interest and within the candidate region is calculated for each of the candidate regions, and the similarity distribution over the entire search region is found. Consequently, it becomes possible to identify, on the basis of the similarity distribution, whether the region of interest is a noise region. For example, a statistic for comparing the minimum value of the similarity and all values of the similarity in the similarity distribution is found. The reliability of the region of interest can be assessed by comparing the statistic and a threshold value.

Description

Ultrasonic imaging apparatus, ultrasonic imaging method, and ultrasonic imaging program

The present invention relates to an ultrasonic imaging method and an ultrasonic imaging apparatus that can clearly identify a tissue boundary when imaging a living body using ultrasonic waves.

In an ultrasonic imaging apparatus used for medical image diagnosis, for example, as described in Patent Document 1, an elastic modulus distribution of a tissue is estimated based on a change amount of a small area of a diagnostic moving image (B-mode image). Thus, a method of displaying the hardness after converting it into a color map is known. However, for example, in the case of the peripheral part of a tumor, the acoustic impedance and the elastic modulus may not differ greatly with respect to the surrounding tissue. I cannot grasp the boundary with other organizations.

Therefore, the technique described in Patent Document 2 proposes a technique that makes it possible to identify a tissue boundary in which acoustic impedance and elastic modulus are not significantly different from the surroundings by creating a scalar field image directly from a motion vector of a diagnostic moving image. Has been.

JP 2004-135929 A JP 2008-79792 A

In the technique disclosed in Patent Document 2, motion vector estimation is performed by performing block matching processing on two diagnostic image data, but an error vector is generated due to the influence of noise in the image data during the estimation processing. To do. Therefore, there has been a problem that the degree of discrimination of the boundary deteriorates. Particularly in the penetration limit region where the echo signal is weak, the vector estimation accuracy is greatly degraded.

An object of the present invention is to provide an ultrasonic imaging apparatus capable of discriminating a noise region where an echo signal is weak.

In order to achieve the above object, according to the first aspect of the present invention, the following ultrasonic imaging apparatus is provided. That is, the ultrasonic imaging apparatus of the present invention includes a transmission unit that transmits ultrasonic waves toward a target, a reception unit that receives ultrasonic waves coming from the target, and a reception signal of the reception unit to process two or more frames. And a processing unit for generating an image. The processing unit sets a region of interest at a predetermined position or a position received from the operator by using one frame of the generated two or more frames as a reference frame. Another frame is set as a comparison frame, a search area wider than the region of interest is set at a predetermined position or a position received from the operator, and a plurality of candidates for the movement destination of the region of interest within the search region Set candidate areas for. The similarity between the image characteristic values in the region of interest and the candidate region is calculated for each candidate region, and the similarity distribution over the entire search region is obtained. This makes it possible to determine whether the region of interest is a noise region based on the similarity distribution.

For example, the processing unit obtains a statistic that compares the minimum value of the similarity and the overall value of the similarity in the similarity distribution, and determines the reliability of the region of interest based on the statistic. Specifically, for example, the processing unit determines the reliability of the region of interest by calculating the above-described statistic using the minimum value, average value, and standard deviation of the similarity, and comparing the calculated statistic with a threshold value. Is possible.

For example, the processing unit can generate a vector connecting the position corresponding to the region of interest in the comparison frame and the position of the candidate region having the minimum similarity, and the vector for the region of interest determined to have low reliability. Is replaced with zero or a predetermined vector. Thereby, an error vector can be removed and the accuracy of the vector can be improved.

For example, the processing unit can calculate the average, minimum value, and standard deviation of the similarity for the similarity distribution, and use the degree of separation obtained by dividing the difference between the average and the minimum value by the standard deviation as the statistic. Further, for example, it is also possible to calculate the average and standard deviation of the similarity for the similarity distribution, and use the coefficient of variation obtained by dividing the standard deviation by the average as the statistic.

Threshold values to be compared with the above-mentioned statistics can be obtained as follows. For example, a plurality of regions of interest are set, a statistic is obtained for each, a histogram distribution indicating the frequency of the obtained statistic value is obtained, and the median, average value, or histogram distribution in the histogram distribution is a plurality of mountain shapes Is used, the statistic of the minimum value of the valley between the mountains is used as the threshold value.

It is also possible to generate a smoothed similarity distribution obtained by smoothing the similarity distribution, and obtain a difference similarity distribution obtained by subtracting the smoothed similarity distribution from the similarity distribution before the smoothing process. Thereby, the fluctuation | variation of the similarity by noise can be removed from similarity distribution.

The smoothing process uses, for example, a method in which a filter having a predetermined size is set in the similarity distribution and the process of smoothing the distribution in the filter is repeated while moving the filter by a predetermined amount. The size of the filter can be determined as follows. A vector connecting the position corresponding to the region of interest in the comparison frame and the position of the candidate region having the minimum similarity in the similarity distribution before the smoothing process is generated in advance for a plurality of regions of interest. Of the generated vectors, the maximum vector length is set as the filter size.

It is also possible to obtain the boundary of the target by filtering the similarity distribution with a Laplacian filter to create a contour enhancement distribution and extracting continuous contour lines in the contour enhancement distribution.

It is also possible to determine the degree of tumor invasion by configuring the processing section as follows. That is, the processing unit obtains a similarity distribution for a region of interest set near the living body tumor boundary, generates a similarity distribution image having the similarity as an image characteristic value, and corresponds to the region of interest on the similarity distribution image. A first processing means for setting a one-dimensional area of a predetermined length in a plurality of different directions centered on a position; a second processing means for calculating the sum of similarities in the one-dimensional area for each set direction; and similarity Third processing means for calculating the ratio between the similarity sum in the direction in which the sum is minimum and the similarity sum of the one-dimensional area in the direction orthogonal thereto, and fourth processing for determining the degree of tumor invasion based on the ratio Means.

When the ratio calculated by the third processing means is smaller than a predetermined value set in advance, it is possible to obtain the boundary line by determining that the target pixel is a point constituting the boundary line.

Further, according to the second aspect of the present invention, the following ultrasonic imaging method is provided. That is, an ultrasonic wave is transmitted toward the target, and a reception signal obtained by receiving the ultrasonic wave coming from the target is processed to generate an image of two or more frames, and a reference frame and a comparison frame are selected from the image. A region of interest is set in the reference frame, a search region wider than the region of interest is set in the comparison frame, a plurality of candidate regions that are candidates for movement of the region of interest are set in the search region, The ultrasonic imaging method calculates the similarity of the image characteristic values in the candidate area for each candidate area and obtains the similarity distribution over the entire search area.

Also, according to the third aspect of the present invention, the following ultrasound imaging program is provided. That is, a first step of selecting a reference frame and a comparison frame from two or more frames of ultrasound images, a second step of setting a region of interest in the reference frame, and a search region wider than the region of interest in the comparison frame The third step of setting a plurality of candidate areas that are candidates for the movement destination of the region of interest in the search area, and calculating the similarity between the image characteristic values in the region of interest and the candidate region for each candidate region And a fourth step of obtaining a similarity distribution over the entire search region.

According to the present invention, it is possible to determine whether the region of interest is a noisy region based on the similarity distribution. As a result, the generation of error vectors is suppressed, and highly accurate vector estimation is possible even in the penetration limit region. The accuracy of the scalar field image converted from the estimated motion vector field is improved, and more appropriate boundary detection is possible.

1 is a block diagram showing an example system configuration of an ultrasound imaging apparatus according to Embodiment 1. FIG. 5 is a flowchart showing a processing procedure for image generation by the ultrasonic imaging apparatus according to the first embodiment. The flowchart which shows the detail of the block matching process of step 24 of FIG. The figure explaining the block matching process of step 24 of FIG. 2 with the phantom of a two-layer structure. (A) The figure which shows the example of a B mode image produced | generated with the ultrasonic imaging apparatus of Embodiment 1, (b) The figure which shows the example of a motion vector distribution image produced | generated with the ultrasonic imaging apparatus of Embodiment 1. FIG. FIG. 5A is a diagram showing an example of an SAD distribution image obtained by setting the ROI at the position (3) in FIG. 5B, and FIG. 5B is obtained by setting the ROI at the position (5) in FIG. The figure which shows the example of a SAD distribution image, (c) is the histogram of the SAD value shown to Fig. (A), (d) is the histogram of the SAD value shown to Fig. (B). FIG. 4 is an explanatory diagram showing the definition of the degree of separation on the histogram of SAD values in the first embodiment. The flowchart which shows the detail of a process in the case of performing the process of step 25 of FIG. 2 using a degree of separation. The figure which shows the example which imaged distribution of the separation degree calculated | required by the process of FIG. The flowchart which shows the detail of a process in the case of performing the process of step 25 of FIG. 2 using a variation coefficient. The figure which shows the example which imaged the distribution of the variation coefficient calculated | required by the process of FIG. (A) The figure which shows the example of a motion vector distribution image produced | generated by step 24 of FIG. 2, (b). The figure which shows the example of a vector distribution image which removed the false vector at step 25 of FIG. 10 is a flowchart of processing for removing noise when calculating the SAD distribution according to the second embodiment. (A) The figure which shows the example of the SAD distribution image before noise removal of Embodiment 2, (b) The figure which shows the example of the SAD distribution image obtained by smoothing (LPF) the SAD distribution image, (c) Noise The figure which shows the example of the SAD distribution image after removal. 9 is a flowchart illustrating processing for obtaining an infiltration degree according to the third embodiment. (A)-(h) Explanatory drawing which shows the area | region selection pattern of SAD distribution used by the process of FIG. (A) is a diagram showing an SAD distribution image obtained by setting the ROI at the position (1) in FIG. 5 (b), and (b) is obtained by setting the ROI at the position (2) in FIG. 5 (b). The figure which shows the SAD distribution image, (c) is a figure which shows the image which applied the Laplacian filter to the SAD value distribution shown in FIG. (A), (d) is the Laplacian filter in the SAD value distribution shown in FIG. The figure which shows the image which applied.

An ultrasonic imaging apparatus according to an embodiment of the present invention will be described below.
(Embodiment 1)
FIG. 1 shows a system configuration of the ultrasonic imaging apparatus of the present embodiment. This apparatus has an ultrasonic boundary detection function. As shown in FIG. 1, this apparatus includes an ultrasonic probe (probe) 1, a user interface 2, a transmission beam former 3, a control system 4, a transmission / reception changeover switch 5, a reception beam former 6, and an envelope detection unit 7. , A scan converter 8, a processing unit 10, a parameter setting unit 11, a combining unit 12, and a display unit 13.

The ultrasonic probe 1 in which ultrasonic elements are arranged one-dimensionally transmits an ultrasonic beam (ultrasonic pulse) to a living body and receives an echo signal (received signal) reflected from the living body. Under the control of the control system 4, a transmission signal having a delay time adjusted to the transmission focus is output by the transmission beamformer 3 and sent to the ultrasonic probe 1 via the transmission / reception changeover switch 5. The ultrasonic beam reflected or scattered in the living body and returned to the ultrasonic probe 1 is converted into an electric signal by the ultrasonic probe 1 and received by the receiving beam former 6 via the transmission / reception switch 5. Sent as.

The receive beamformer 6 is a complex beamformer that mixes two received signals that are 90 degrees out of phase, and performs dynamic focus that adjusts the delay time according to the reception timing under the control of the control system 4. Outputs real and imaginary RF signals. The RF signal is detected by the envelope detector 7 and then converted into a video signal, which is input to the scan converter 8 and converted into image data (B-mode image data). The configuration described above is the same as the configuration of a known ultrasonic imaging apparatus.

In the apparatus of the present invention, the ultrasonic boundary detection process is realized by the processing unit 10. The processing unit 10 includes a CPU 10a and a memory 10b. When the CPU 10a executes a program stored in the memory 10b in advance, the following processing is performed to detect the boundary of the subject tissue. That is, based on the image data of two or more frames output from the scan converter 8, the processing unit 10 first creates a motion vector field. Next, the generated motion vector field is converted into a scalar field. The original image data and the corresponding motion vector field or scalar field are combined by the combining unit 12 and then displayed on the display unit 13.

The parameter setting unit 11 performs parameter setting for signal processing in the processing unit 10 and selection setting of a display image in the synthesis unit 12. These parameters are input from the user interface 2 by an operator (device operator). As parameters for signal processing, for example, setting of a region of interest on a desired frame m and setting of a search region on a frame m + Δ different from the frame m can be received from an operator. As the selection setting of the display image, for example, the selection setting of whether the original image and the vector field image (or scalar image) are combined into one image and displayed on the display, or two or more moving images are displayed side by side. Can be received from the operator.

FIG. 2 shows a flowchart of an example of boundary detection processing and image processing in the processing unit 10 and the combining unit 12 of the present invention. The processing unit 10 first acquires a measurement signal from the scan converter 8 and performs normal signal processing to create a B-mode moving image (steps 21 and 22). Next, two frames of a desired frame and a frame having a different time are extracted from the B-mode moving image (step 23). For example, the desired frame and the next frame are extracted. Then, a motion vector field is calculated from the two frames (step 24). The motion vector field is calculated based on the block matching method. A noise removal process is performed on the calculated motion vector field (step 25), and the noise-removed motion vector field is converted to a scalar field (step 26). Then, the scalar field image and the motion vector field image or B-mode image are combined and displayed, and the processing for one image is completed (step 27). By selecting different frames sequentially in time series as desired frames in step 23 and repeating the processing of steps 21 to 27 described above, and continuously displaying the composite image, it is possible to display a moving image of the composite image.

FIG. 3 is a flowchart showing detailed processing in step 24, and FIG. 4 is a diagram for explaining block matching processing in step 24. The block matching process for calculating the motion vector field in step 24 will be specifically described with reference to FIGS. Here, it is assumed that the mth and m + Δth frames are selected in step 23. For example, Δ = 1 frame. First, in the frame m, the processing unit 10 sets a region of interest ROI (region of interest: reference block) 31 having a predetermined number of pixels N as shown in FIG. 4 (step 51). The luminance distribution of the pixels included in the ROI 31 is represented as P _m (i ₀ , j ₀ ). i ₀ and j ₀ indicate the position of the pixel in the ROI 31. Next, the processing unit 10 sets a search area 32 of a predetermined size at a position corresponding to the ROI 31 of the frame m and its vicinity at the frame m + Δth (step 52). Here, the setting of the ROI 31 will be described with respect to a configuration in which the processing unit 10 sequentially sets the ROI 31 over the entire image of the frame m and sets a search area 32 of a predetermined size centered on the ROI 31. 10 is a ROI 31 having a predetermined position and size, and a search region 32 having a predetermined size is set in the vicinity thereof, and a region of interest (ROI) and a search region received from an operator in the parameter setting unit 11 In addition, the processing unit 10 can set the ROI 31 and the search area 32.

The search area 32 is divided into a plurality of movement candidate areas 33 having the same size as the ROI 31. The processing unit 10 calculates the movement candidate area 33 having the highest similarity to the luminance of the ROI 31 and selects it as the movement destination area. As an index representing the degree of similarity, a sum of absolute differences, a mean square, a cross-correlation value, or the like can be used. Here, the case where the sum of absolute differences is used as an example will be described below.

The luminance distribution of the pixels included in the movement candidate area 33 in the search area 32 is represented as P _{m + Δ} (i, j). i and j indicate the position of the pixel in the movement candidate area 33. The processing unit 10 calculates the sum of absolute differences SAD (sum of absolute difference) between the luminance distribution P _{m + Δ} (i, j) of the pixel of the ROI 31 and the luminance distribution P _m (i ₀ , j ₀ ) of the movement candidate region 33. (Step 53). Here, SAD is defined by the following equation (1).

The processing unit 10 obtains the SAD value with the ROI 31 for all the movement candidate areas 33 in the search area 32, and determines the movement candidate area 33 having the smallest SAD value in the obtained SAD distribution as the movement destination area. Then, a motion vector connecting the position of the ROI 31 and the position of the movement candidate area 33 with the minimum SAD value is determined (step 54).

Then, the processing unit 10 determines the motion vector for the entire image of the frame m by repeating the above process while moving the ROI 31 over the entire image of the frame m (step 55). A motion vector field (motion vector distribution image) is obtained by generating an image in which the determined vector is indicated by an arrow, for example.

FIGS. 5A and 5B are examples of B-mode images and motion vector distribution images obtained by the above processing. The B-mode image in FIG. 5A is obtained by superposing the gel

base material phantoms

41 and 42 in two layers and fixing the ultrasonic probe to the upper phantom 41 and moving it laterally. FIG. 5B shows the motion vector obtained by the block matching process (step 24) with the next frame (frame m + Δ, Δ = 1 frame), where the B-mode image in FIG.

As shown in FIG. 5B, the region of the upper 上側 corresponding to the upper phantom 41 to which the ultrasonic probe 1 is fixed is relatively stationary, and the lower phantom 42 is a vector field indicating lateral movement. (Horizontal arrow). However, in the region of the lower 領域 of the lower phantom 42 in FIG. 5B, the arrow points obliquely upward, the direction is not constant, and there is a phenomenon that the motion vector is disturbed. This phenomenon is caused by a decrease in the S / N ratio (SNR) of detection sensitivity as the distance from the probe 1 increases, and indicates a penetration limit. That is, an erroneous vector is generated in a low SNR region far from the probe 1.

The noise removal process of the motion vector field in step 25 will be described with reference to FIG.

FIGS. 6A and 6B show the respective movement candidate areas 33 of the search area 32 of the next frame obtained by setting the ROI 31 at the positions (3) and (5) in FIG. FIG. 6 is a diagram (SAD distribution diagram) showing SAD values as densities of respective movement candidate regions 33; FIGS. 6A and 6B each show a search area 32 as a whole, and the search area 32 is divided into 21 × 21 movement candidate areas 33. The block size of the movement candidate area 33 is 30 × 30 pixels, the search area 32 is 50 × 50 pixels, and the movement candidate area 33 is moved pixel by pixel within the search area 32. The search area 32 is set so that the position of the movement candidate area 33 corresponding to the position of the ROI 31 is located at the center of the search area 32.

In the SAD distribution of FIG. 6A, the SAD value of the movement candidate region 33 at a position shifted to the right side from the center of the search region 32 is minimum. Therefore, it can be seen that a right lateral vector is determined in step 24 for the position (3) in FIG. As can be seen from FIG. 5 (b), the position (3) is slightly below the boundary between the two-layer phantoms 41 and 42 (the phantom 42 side), and a vector in the horizontal direction is displayed.

If attention is paid to the spatial distribution of the SAD values in FIG. 6A, the SAD values are distributed around the movement candidate region 33 having the smallest SAD value in the horizontal direction, that is, in the direction along the boundary between the two-

layer phantoms

41 and 42. It can be seen that a small region (SAD value valley) is formed. This phenomenon is that, as shown in FIG. 6B, the boundary can be directly detected from the SAD distribution of FIG. 6A without creating all the motion vector fields while moving the ROI 31 to the entire region of the frame m. It suggests.

On the other hand, since the SAD value distribution of FIG. 6B is the position (5) in the region where the motion vector is disturbed in FIG. 5B, the movement candidate region 33 with the smallest SAD value is the noise. It can be seen that the value is uniformly spread over a wide area near the probe 1 where the value is small, and the motion vector is upward and easily disturbed. It can also be seen that the region with the small SAD value (the valley of the SAD value) that should be formed around the movement candidate region 33 with the smallest SAD value is buried in the entire noise fluctuation and cannot be recognized.

In the present invention, as shown in FIG. 6A, in the SAD distribution image in which the ROI 31 is set at the position (3) where the noise is small, the SAD value in which the movement candidate region 33 having the smallest SAD value is clearly smaller than the surrounding region. Therefore, in the SAD distribution image of the search area 32 in which the ROI 31 is set at the noisy position (5) as shown in FIG. 6B, the movement candidate area 33 having the smallest SAD value is obtained. However, the noisy ROI 31 is determined using the phenomenon that it cannot be clearly recognized. For the noisy ROI 31, a process for removing the corresponding motion vector is performed.

For the above-described determination, the processing unit 10 determines, for example, from the SAD distribution of the search region 32 as shown in FIGS. 6A and 6B when the ROI 31 is set at a predetermined position in step 25 of FIG. Histograms showing the distribution of SAD values are created as shown in FIGS. When the ROI 31 is set at the position (3) in FIG. 5B, the SAD minimum value is sufficiently separated from the SAD value having a high frequency in the histogram distribution as shown in FIG. 6C. . That is, the minimum SAD value and the range of SAD values with high frequency are sufficiently separated. On the other hand, the histogram when the ROI 31 is set at the position (5) in FIG. 5B does not have much frequency difference with respect to the SAD value distribution as shown in FIG. 6D, and the histogram distribution is broad. It has spread. For this reason, the SAD minimum value is included in the range of the SAD value having a high frequency, and the separation between the minimum SAD value and the range of the SAD value having a high frequency is insufficient.

From such characteristics, by comparing the SAD minimum value of the histogram of the SAD distribution in the search region 32 with the SAD value having a high frequency, the reliability of the signal of the ROI 31 corresponding to this search region 32 (less noise) ) And the reliability of the motion vector determined in the search area 32 can be determined. As a result, a region with low reliability can be determined as a low SNR region, and the reliability of the corresponding motion vector can also be determined.

In the present invention, an index is used to determine the degree of separation between the SAD minimum value and the frequent SAD value in the histogram of the SAD value distribution. First, a processing method in the case where the separability parameter is used as an index will be described first. FIG. 7 shows the concept of definition of the degree of separation. The degree of separation is a value corresponding to the distance between the distribution average of the histogram and the minimum value, and is defined by the following equation (2).

In Formula (2), in order to avoid the influence by the difference in distribution, normalization is performed with the standard deviation.

FIG. 8 shows a flow of processing for discriminating a noisy region when the degree of separation is used as an index and removing a motion vector. This process specifically shows the process of step 25 in FIG. 2, and is performed for all ROIs 31 set in frame m in step 51 of FIG. The processing unit 10 first determines the target ROI 31 (step 81), and the SAD value of all the movement candidate regions 33 in the search region 32 set and calculated in

steps

52 and 53 in FIG. 3 for the determined ROI 31. Is used to calculate the mean, minimum value, and standard deviation of the SAD values according to statistical processing (step 82). Then, the degree of separation defined by the above equation (2) is obtained (step 83). This is repeated for all ROIs 31 (step 84). Since the ROI 31 whose degree of separation obtained in step 84 is lower than a predetermined value is a region where the noise is large and the reliability of the motion vector is low, the motion vector obtained in step 54 in FIG. 3 is replaced with 0 (step 85). ). As a result, a region with low reliability can be determined from the motion vector image, and the vector (erroneous vector) can be removed.

As the predetermined value for determining whether or not the degree of separation is low in Step 85, a predetermined threshold value or the median value of the distribution of the degree of separation obtained for all ROIs 31 in Step 84 can be used. Alternatively, when a histogram of the obtained degree of separation is generated and a plurality of frequency peaks are formed, between the mountain located on the side with the lowest degree of separation and the mountain located on the side with the higher degree of separation. It is also possible to use the degree of separation of the valley positions as a threshold value.

FIG. 9 shows an image of the degree of separation obtained for all ROIs 31 in step 83. In FIG. 9, 33 × 51 ROIs 31 are set in the frame m, and the separation degree of each ROI 31 is indicated by the density. As shown in FIG. 9, the degree of separation is low in the low SNR region below the frame m, and it can be seen that the degree of separation reflects the reliability of motion vector estimation.

In the process of FIG. 8, the degree of separation is used. However, as an index for determining the degree of separation between the SAD minimum value and the SAD value having a high frequency, another index can be used. For example, a coefficient of variation can be used. The coefficient of variation is defined by the following equation, and is a statistic obtained by standardizing standard deviations on the average, and indicates the magnitude of distribution variation (that is, difficulty in separating minimum values).

FIG. 10 shows a flow of processing for removing a vector in a noisy area when the coefficient of variation is used as an index. In this process, the target ROI 31 is determined in the same manner as in the processing flow of FIG. 8 (step 81), and all the movement candidates in the search area 32 set and calculated in

steps

52 and 53 of FIG. Using the SAD values in the region 33, the average of the SAD values and the standard deviation are calculated according to statistical processing (step 101). Then, the coefficient of variation defined by the above equation (3) is obtained (step 102). This is repeated for all ROIs 31 (step 84). For the ROI 31 whose coefficient of variation obtained in step 102 is larger than a predetermined value, the motion vector obtained in step 54 of FIG. 3 is replaced with 0 (step 85). As a result, it is possible to determine the ROI 31 having a large noise, determine an unreliable region from the motion vector image, and remove the vector (false vector).

As the predetermined value for determining whether or not the variation coefficient is high in step 85, a predetermined threshold value or the median value of the distribution of variation coefficients obtained for all ROIs 31 in step 84 can be used. Alternatively, when a histogram of the obtained variation coefficient is generated and the frequency exhibits two peaks, it is also effective to adopt the minimum value of the valley between the two peaks as the predetermined value.

FIG. 11 shows the variation coefficient obtained for all ROIs 31 in step 102 as an image. In FIG. 11, the coefficient of variation of each ROI 31 is shown by the concentration. As shown in FIG. 11, the coefficient of variation increases in the low SNR region at the bottom of the frame m, and it can be seen that the coefficient of variation reflects the reliability of motion vector estimation.

FIGS. 12A and 12B show examples of motion vector distribution images before and after removing an erroneous vector. FIG. 12A shows the same motion vector field before error vector removal as FIG. 4B, and FIG. 12B obtains the coefficient of variation distribution from the SAD distribution by the processing of FIG. An ROI having a median value as a threshold and a coefficient of variation larger than that is determined to have low reliability, and the motion vector is set to 0 (stationary state). When comparing FIGS. 12A and 12B, the lower region where the motion vector is disturbed is clearly removed and replaced with a stationary state. In other words, the lower region can be determined as a penetration region (that is, a region with low reliability) where an ultrasonic echo cannot be accurately obtained.

If the error vector of the motion vector distribution is removed by the processing of FIG. 8 and FIG. 10 described above, the motion vector distribution is converted into a scalar distribution by

steps

26 and 27 in FIG. The distribution image (or B-mode image) is synthesized and displayed.

In the processing of FIGS. 8 and 10, the motion vector in the region with low reliability is removed to make it stationary. However, the present invention is not limited to this processing method. For example, instead of setting the motion vector to a static state, it is possible to use a processing method that maintains the state of the motion vector obtained previously for the same region as it is.

(Embodiment 2)
In the first embodiment, after obtaining the SAD distribution, whether or not the ROI 31 is in the low SNR region is determined by the SAD distribution, and in the case of the low SNR region, the motion vector is removed. When calculating the SAD distribution, noise is removed, and a highly reliable motion vector is obtained using the SAD distribution with high detection sensitivity after noise removal.

FIG. 13 is a processing flow for removing noise during the calculation of the SAD distribution of the second embodiment. In the processing of FIG. 13, noise removal processing (steps 132 and 133) is added to the SAD calculation processing of steps 51 to 54 of FIG. 3 of the first embodiment. 14A, 14B, and 14C show examples of the SAD distribution at each processing stage in FIG.

As shown in FIG. 13, the processing unit 10 first performs the same processing as steps 51 to 53 in FIG. 3 of the first embodiment to obtain the SAD value distribution. An example of the obtained SAD distribution image is shown in FIG. The ROI 31 is at the position (4) in FIG. 5B, and the SAD value of the upper movement candidate area 33 is affected by noise even though the phantom 42 originally moves relatively in the horizontal direction. Since the motion vector is determined as it is, erroneous detection of the vector occurs. In order to avoid such erroneous detection, the processing unit 10 performs a smoothing process (low-pass filter LPF (low pass filter) process) on the SAD distribution image obtained in steps 51 to 53 (step 132). In the smoothing process, for example, a filter having a predetermined size is applied to the SAD distribution image, and the high frequency component of the SAD distribution in the filter is cut and smoothed. This process is repeated while moving the filter by a predetermined amount. As described above, by smoothing the SAD value distribution image, the change in the SAD value due to the movement of the phantom 42 is removed because it is steep, whereas the change in the AD value due to noise is gradual and can be extracted. . An SAD value distribution image obtained by the smoothing process is shown in FIG.

Next, the difference between the original SAD value distribution at step 53 and the smoothed SAD value distribution at step 132 is obtained (step 133). Thereby, the original SAD value distribution by the movement of the phantom from which the fluctuation of the SAD value due to noise is removed can be obtained. The obtained distribution is shown in FIG. Using the obtained SAD value distribution, step 54 of FIG. 3 is performed, the movement candidate area 33 having the smallest SAD value is determined as a movement destination, and a motion vector is determined. After the motion vector is determined, a motion vector distribution image is generated by the process of step 56 in FIG. 3 of the first embodiment. Further, in step 25 of FIG. 2 in the first embodiment, it is possible to further perform processing such as removing a vector with low reliability of the motion vector distribution.

As described above, since the motion vector can be determined using the SAD value distribution from which the SAD value fluctuation due to noise is removed by the processing of the second embodiment, the reliability of the motion vector can be improved.

In the above smoothing process, LPF is used. However, the present invention is not limited to this. If the spatial frequency of the distribution of SAD values due to movement of the subject (phantom) is high (that is, a more complicated shape), the band pass is used. It is also effective to apply a filter.

The size of one side of the filter used in the filter processing in step 132 can be determined as follows. That is, for the SAD distribution that has not been smoothed, the motion vector field is created in advance by performing step 54 in FIG. Set as the size of one side.

(Embodiment 3)
Next, as a third embodiment, a processing method for directly determining a tissue boundary using the SAD value distribution obtained in step 53 of FIG. 3 of the first embodiment and determining an infiltration degree of a living tumor with respect to a normal tissue Will be described. In the third embodiment, the movement candidate area 33 of the search area 32 is simply referred to as an area 33. The ROI 31 is also referred to as a target pixel.

The region along the boundary of the tissue of the subject shows high brightness in the B-mode image because the tissue similarity is high. For this reason, the SAD has a feature that the region 33 along the boundary of the tissue of the subject has a smaller value than the region 33 along the orthogonal direction of the boundary. On the other hand, when the invasion of the living tumor progresses, the boundary becomes unclear, so the SAD value of the region 33 along the boundary increases. Using this, the degree of infiltration is determined.

FIG. 15 shows a processing flow of the processing unit 10 of the third embodiment. FIGS. 16A to 16H show eight patterns of the target direction and the region 33 selected on the SAD value distribution image correspondingly.

First, the processing unit 10 sets a pixel of interest (ROI) 31 at the boundary position of the tissue to be investigated in the desired frame m of the B-mode image, sets a search area 32 in the frame m + Δ, and calculates the SAD value distribution of the search area 32. Obtain (step 151). The frame selection method and the SAD value distribution calculation method are performed in the same manner as steps 21 to 23 in FIG. 2 and steps 51 to 53 in FIG.

In this SAD value distribution, as shown in FIG. 16A, an area 33 passing through the center of the search area 32 and positioned along a predetermined target direction (horizontal direction) 151 is selected (step 63), and the selected area 33 is selected. The sum of the SAD values is obtained (step 64). Similarly, the region 33 positioned along the orthogonal direction (vertical direction) 152 of the target direction 151 is selected, and the sum of the SAD values of the selected region 33 is also obtained.

The processes in

steps

63 and 64 are repeated until all the eight patterns shown in FIGS. 16A to 16H are performed (step 62). In the pattern of FIG. 16B, the sum of the SAD values of the region 33 located along a predetermined target direction (direction inclined about 30 ° counterclockwise with respect to the horizontal direction) 151 is obtained, and the orthogonal direction 152 is obtained. The sum of the SAD values of the region 33 located along is obtained.

In the patterns of FIGS. 16C to 16H, the target directions 151 of about 45 °, about 60 °, 90 °, about 120 °, about 135 °, and about 150 ° counterclockwise with respect to the horizontal direction; The sum of the SAD values of the region 33 located along the orthogonal direction 152 is obtained.

The target direction 151 that minimizes the sum of the calculated SAD values of each target direction 151 is selected (step 65). The direction of the selected target direction 151 is the direction of the tissue boundary. Thereby, the boundary can be directly detected without obtaining a motion vector.

Next, a direction 152 orthogonal to the selected target direction 151 is selected (step 66). The ratio of the sum of the SAD values in the selected target direction 151 and the SAD sum in the direction 152 orthogonal to the selected target direction 151 (the SAD value sum in the target direction / the SAD value sum in the orthogonal direction) is calculated (step 67).

When the degree of infiltration is low and the boundary is clear, the SAD sum in the boundary direction (selected target direction 151) is small and the SAD sum in the orthogonal direction 152 is large, so a small value is obtained as the ratio. On the other hand, as the degree of infiltration increases and the boundary becomes unclear, the SAD sum in the boundary direction (selected target direction 151) increases, so the ratio increases. Therefore, the degree of infiltration can be evaluated using the ratio as a parameter. Specifically, for example, the degree of infiltration is determined by comparing a plurality of predetermined reference values and ratios, and the determination result is displayed.

When the ratio is smaller than a predetermined value, the target pixel (ROI 31) can be identified as a point constituting a boundary line and the boundary can be displayed.

Note that a direction-dependent filter can be used when calculating the sum of the SAD values in each direction. The direction-dependent filter is a filter having a function of determining a direction in which the density change in the one-dimensional direction is the smallest in the filter range (search region 32) of the processing pixel.

FIGS. 16A to 16H show region selection patterns in the target direction 151 and the orthogonal direction 152 in the search region 32 composed of the 5 × 5 region 33 for easy illustration. In the processing, an area selection pattern is set corresponding to the number of areas 33 in the search area 32.

(Embodiment 4)
As a fourth embodiment, another method for directly detecting a boundary without obtaining a motion vector from the SAD value distribution in the search region 32 will be described. Here, a Laplacian filter that performs enhancement processing corresponding to second-order differentiation is applied.

FIGS. 17A and 17B show the SAD distribution of the search region 32 by setting the ROI 31 at the positions (1) and (2) in FIG. 5B. The position (1) is a position inside the phantom 41 that is relatively stationary with respect to the probe 1. The position (2) is located in the vicinity of the boundary between the phantom 41 and the phantom 42 moving laterally relative thereto. The processing for obtaining the SAD distribution at the positions (1) and (2) is performed in the same manner as steps 21 to 23 in FIG. 2 and steps 51 to 53 in FIG. When a Laplacian filter that performs spatial quadratic differentiation is applied to the obtained SAD distribution image (FIGS. 17A and 17B), a portion where the fluctuation of the SAD value is greatly emphasized due to the edge enhancement effect. Images of 17 (c) and (d) are obtained, respectively.

When there is a boundary in the search region 32, as shown in FIG. 17D, a region with a large variation in SAD value along the boundary is emphasized and a contour enhancement distribution extracted in a streak shape is generated. The Therefore, by extracting and displaying a continuous contour region (movement candidate region 33) in the contour enhancement distribution by binarization processing, the boundary can be detected from the Laplacian image of the SAD value. On the other hand, when there is no boundary in the search region 32, as shown in FIG. 17C, the region where the SAD value varies greatly is only the central region (the movement candidate region 33 corresponding to the position of the ROI 31). Yes, it is isolated from the surroundings and does not become a continuous outline. Therefore, it can be seen that the boundary is not at this position.

According to the above processing, the boundary can be extracted directly by performing Laplacian processing on the SAD distribution. Therefore, steps 54 and 55 in FIG. 3 for determining and imaging the motion vector of the first embodiment, and steps 25 to 55 in FIG. 2 in which the motion vector is denoised and converted into a scalar distribution to estimate the boundary. Since 26 can be omitted, the amount of processing can be greatly reduced.

The present invention can be applied to medical ultrasonic diagnostic / treatment apparatuses and apparatuses that measure distortion and deviation using electromagnetic waves including ultrasonic waves in general.

1: Ultrasonic probe (probe) 2: User interface 3: Transmission beamformer 4: Control system 5: Transmission / reception changeover switch 6: Reception beamformer 7: Envelope detector 8: Scan converter, 10: processing unit, 10a: CPU, 10b: memory, 11: parameter setting unit, 12: synthesis unit, 13: display unit.

Claims

A transmission unit that transmits ultrasonic waves toward the target; a reception unit that receives ultrasonic waves coming from the target; and a processing unit that generates an image of two or more frames by processing a reception signal of the reception unit. And
The processing unit sets a region of interest at a predetermined position or a position received from an operator from one of the generated two or more frames as a reference frame, and sets another frame as a reference frame. As a comparison frame, a search area wider than the region of interest is set to a predetermined position or a position received from an operator, and a plurality of candidate regions that are candidates for the destination of the region of interest are included in the search region. An ultrasound imaging apparatus characterized by setting and calculating a similarity between image characteristic values in the region of interest and the candidate region for each candidate region to obtain a distribution of the similarity over the entire search region .
The ultrasonic imaging apparatus according to claim 1, wherein the processing unit obtains a statistic for comparing the minimum value of the similarity and the overall value of the similarity in the similarity distribution, and uses the statistic to calculate the statistic. An ultrasound imaging apparatus, wherein the reliability of the region of interest is determined.
The ultrasonic imaging apparatus according to claim 2, wherein the processing unit obtains the statistic using the minimum value, average value, and standard deviation of the similarity, and compares the obtained statistic with a threshold value. And determining the reliability of the region of interest.
The ultrasound imaging apparatus according to claim 3, wherein the processing unit generates a vector connecting a position corresponding to the region of interest in the comparison frame and a position of the candidate region having the minimum similarity, An ultrasonic imaging apparatus, wherein the vector of interest is determined to have a low reliability, and the vector is replaced with zero or a predetermined vector.
The ultrasound imaging apparatus according to claim 2, wherein the processing unit calculates an average, a minimum value, and a standard deviation of the similarity for the similarity distribution, and divides a difference between the average and the minimum value by the standard deviation. An ultrasonic imaging apparatus characterized in that the degree of separation obtained is obtained as the statistic.
3. The ultrasound imaging apparatus according to claim 2, wherein the processing unit calculates an average of the similarity and a standard deviation for the similarity distribution, and obtains a coefficient of variation obtained by dividing the standard deviation by the average as the statistic. An ultrasonic imaging apparatus characterized by that.
The ultrasonic imaging apparatus according to claim 3.
The processing unit sets a plurality of regions of interest, obtains the statistic for each, obtains a histogram distribution indicating the frequency of the obtained statistic value,
An ultrasonic imaging apparatus characterized in that, when the median value, average value, or histogram distribution in the histogram distribution has a plurality of mountain shapes, a statistic of a minimum value of a valley between mountains is used as the threshold value. .
The ultrasound imaging apparatus according to claim 1,
The processing unit generates a smoothed similarity distribution obtained by smoothing the similarity distribution, and obtains a difference similarity distribution obtained by subtracting the smoothed similarity distribution from the similarity distribution before the smoothing process. An ultrasonic imaging apparatus characterized by that.
9. The ultrasonic imaging apparatus according to claim 8, wherein the smoothing process includes a process of setting a filter of a predetermined size in the similarity distribution and smoothing the distribution in the filter. It repeats while moving quantitatively,
The processing unit uses a vector connecting a position corresponding to the region of interest in the comparison frame and a position of the candidate region having the minimum similarity in the similarity distribution before the smoothing process as a plurality of the regions of interest. And the maximum vector length among the generated vectors is the size of the filter.
The ultrasonic imaging apparatus according to claim 1, wherein the processing unit creates a contour enhancement distribution by filtering the similarity distribution with a Laplacian filter, and extracts a continuous contour line in the contour enhancement distribution. An ultrasonic imaging apparatus for obtaining a boundary of the object.
The ultrasonic imaging apparatus according to claim 1, wherein the processing unit obtains the similarity distribution for the region of interest set in the vicinity of a living body tumor boundary, and obtains a similarity distribution image having the similarity as an image characteristic value. Generating and setting a one-dimensional region of a predetermined length in a plurality of different directions around the position corresponding to the region of interest on the similarity distribution image;
A second processing means for calculating the sum of the similarities in the one-dimensional region for each set direction;
A third processing means for calculating a ratio between a similarity sum in a direction in which the similarity sum is minimum and a similarity sum of a one-dimensional region in a direction orthogonal thereto;
Fourth processing means for determining the degree of tumor invasion based on the ratio;
An ultrasonic imaging apparatus comprising:
2. The ultrasound imaging apparatus according to claim 1, wherein when the ratio calculated by the third processing unit is smaller than a predetermined value, the processing unit forms a boundary line. 3. An ultrasonic imaging apparatus characterized in that it is determined to be a point.
Transmitting an ultrasonic wave toward the object, processing the reception signal obtained by receiving the ultrasonic wave coming from the object, and generating an image of two or more frames,
Select a reference frame and a comparison frame from the image,
A region of interest is set in the reference frame;
A search area wider than the region of interest is set in the comparison frame, and a plurality of candidate regions that are candidates for movement of the region of interest are set in the search region,
An ultrasonic imaging method, wherein similarity between image characteristic values in the region of interest and the candidate region is calculated for each candidate region, and the similarity distribution over the entire search region is obtained.
On the computer,
A first step of selecting a reference frame and a comparison frame from an ultrasonic image of two or more frames;
A second step of setting a region of interest in the reference frame;
A third step of setting a search region wider than the region of interest in the comparison frame, and setting a plurality of candidate regions that are candidates for the destination of the region of interest in the search region;
A fourth step of calculating a similarity between image characteristic values in the region of interest and the candidate region for each candidate region, and obtaining a distribution of the similarity over the entire search region;
Ultrasound imaging program for running